滚动轴承声发射信号的人工神经网络模式识别技术  

Investigation of artificial neural network pattern recognition of acoustic emission signals for rolling bear

作　　者：于江林[1,2] 余永增[2] 戴光[2] 汪雪[2] 

机构地区：[1]大庆炼化公司机动设备处,黑龙江大庆163411 [2]大庆石油学院机械科学与工程学院,黑龙江大庆163318

出　　处：《大庆石油学院学报》2008年第5期64-66,共3页Journal of Daqing Petroleum Institute

基　　金：黑龙江省自然科学基金项目(E2007-02);黑龙江省教育厅科学技术研究项目(11521004)

摘　　要：根据声发射检测中常用的BP、RBF和PNN神经网络模型,利用声发射在线检测系统对故障滚动轴承进行测试,提取不同故障轴承声发射信号特征参量作为神经网络输入向量,并分别用3种神经网络对滚动轴承故障模式进行识别.结果表明,采用BP神经网络的声发射信号识别技术的正确识别率略低于其余2种的识别率;RBF和PNN网络的分类结果相同,且在分类能力和学习速度方面均优于BP网络.Three neural network models BP,RBF and PNN that are in common use in AE inspection were introduced in the paper.The acoustic emission signals of every fault pattern bears are inspected by AE inline inspecting system.AE characteristic parameters are extracted and put into neural networks finally identified the condition and fault pattern using these three networks.The diagnostic performance indicates that the correct recognition rate of BP network is slightly lower than other two kinds,the classified results of RBF and PNN network are the same,and the classified ability and the study speed of these two both surpass the BP network.

关 键 词：滚动轴承 声发射 非接触 人工神经网络 模式识别 

分 类 号：TH133.3[机械工程—机械制造及自动化] TN911.71[电子电信—通信与信息系统]